diff --git a/bk7231tools/serial/cmd_hl_flash.py b/bk7231tools/serial/cmd_hl_flash.py
index 71a1c21..183b090 100644
--- a/bk7231tools/serial/cmd_hl_flash.py
+++ b/bk7231tools/serial/cmd_hl_flash.py
@@ -138,14 +138,15 @@ def program_flash(
             # write data in 256-byte chunks
             sector_end = (addr & ~0xFFF) + 4096
             while addr & 0xFFF:
-                block = io.read(min(256, sector_end - addr))
+                padding_len = addr & 0xFF
+                block = io.read(min(256 - padding_len, sector_end - addr))
                 block_size = len(block)
                 if not block_size:
                     # writing finished
                     return
                 self.flash_write_bytes(
-                    addr,
-                    block,
+                    addr - padding_len,
+                    padding_len * b"\xFF" + block,
                     crc_check=crc_check,
                     dry_run=dry_run,
                 )
diff --git a/bk7231tools/serial/cmd_ll_flash.py b/bk7231tools/serial/cmd_ll_flash.py
index 8945179..4aa1264 100644
--- a/bk7231tools/serial/cmd_ll_flash.py
+++ b/bk7231tools/serial/cmd_ll_flash.py
@@ -136,6 +136,17 @@ def flash_write_bytes(
         if response.written != len(data):
             raise ValueError(f"Writing failed; wrote only {response.written} bytes")
         if crc_check:
+            if start & 0xFF:
+                self.warn(
+                    f"Start address (0x{start:06X} is not 256 B-aligned, "
+                    f"can't check CRC"
+                )
+                return
+            if len(data) < 256:
+                # start address aligned, but length is less than 256
+                # add necessary padding
+                # (this is an assumption, that the block was erased prior to writing)
+                data += b"\xFF" * (256 - len(data))
             self.check_crc(start, data)
 
     def flash_write_4k(
@@ -145,6 +156,8 @@ def flash_write_4k(
         crc_check: bool = True,
         dry_run: bool = False,
     ) -> None:
+        if start & 0xFFF:
+            raise ValueError(f"Start address (0x{start:06X}) is not 4k-aligned")
         if len(data) > 4096:
             raise ValueError(f"Data too long ({len(data)} > 4096)")
         if len(data) < 4096:
diff --git a/bk7231tools/serial/protocol.py b/bk7231tools/serial/protocol.py
index b2241f3..5d6187d 100644
--- a/bk7231tools/serial/protocol.py
+++ b/bk7231tools/serial/protocol.py
@@ -123,7 +123,7 @@ def command(
                 offset = self.fix_addr(offset)
                 setattr(packet, field, offset)
 
-        self.debug("<- TX:", shorten(str(packet), 64))
+        self.debug("<- TX:", shorten(str(packet), 100))
 
         self.write(self.encode(packet))
         if after_send:
@@ -187,7 +187,7 @@ def command(
                 else:
                     resp = response.hex()
                 raise ValueError(f"Couldn't deserialize response: {resp}")
-            self.debug(f"-> RX ({size}):", shorten(str(response), 64))
+            self.debug(f"-> RX ({size}):", shorten(str(response), 100))
             # check response status code, if available
             if response.STATUS_FIELDS:
                 for field in response.STATUS_FIELDS: